Skip to main content

YOLO Environment Construction

This document describes how to set up a YOLO development and inference environment on NVIDIA Jetson platforms using Ultralytics YOLO with GPU acceleration.


1. System Information

Verify your system environment before installation.

System
Information


2. Preliminary Preparation

Update system packages and ensure pip is available:

sudo apt update
sudo apt install python3-pip -y
sudo pip install -U pip

3. Install Ultralytics YOLO

Install the Ultralytics YOLO framework with export support:

sudo pip3 install ultralytics[export]

Reboot the system after installation:

sudo reboot

4. Configure GPU Acceleration

note

Torch and Torchvision were installed in previous sections. The following steps install additional GPU-related dependencies.

4.1 Torch

sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.5.0a0+872d972e41.nv24.08-cp310-cp310-linux_aarch64.whl

4.2 Torchvision

sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.20.0a0+afc54f7-cp310-cp310-linux_aarch64.whl

4.3 cuSPARSELt

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/arm64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install libcusparselt0 libcusparselt-dev

4.4 ONNX Runtime GPU

sudo pip3 install https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.20.0-cp310-cp310-linux_aarch64.whl
warning

onnxruntime-gpu requires a specific NumPy version.

Install the compatible NumPy version:

sudo pip3 install numpy==1.23.5

5. Verify the Installation

Verify Ultralytics

python3 -c "import ultralytics; print(ultralytics.__version__)"

Verify Torch

python3 -c "import torch; print(torch.__version__); print(torch.cuda.is_available())"

Verify Torchvision

python3 -c "import torchvision; print(torchvision.__version__)"

Verify NumPy

python3 -c "import numpy; print(numpy.__version__)"

Verification
Output


6. Common Errors

6.1 Cannot Uninstall sympy

Error: Unable to uninstall sympy

Sympy
Error

Solution:

sudo apt remove python3-sympy -y

Reinstall PyTorch afterward if required.


6.2 CSI Camera Cannot Be Called

If the CSI camera does not work with YOLO inference:

  • Rebuild OpenCV from source
  • Ensure CUDA and GStreamer support are enabled
  • Remove old OpenCV versions before reinstalling

This resolves most CSI camera access issues on Jetson platforms.


References


Maintained by HemiHex for Jetson-based advanced vision and YOLO workflows.